1. Field of the Invention
The present invention pertains to the art of imaging and, more particularly, to an image collection and processing system that passively detects skin through spectral measurement of an acquired image.
2. Discussion of the Prior Art
There is a constant need, particularly in the military and security fields, for advanced technologies that provide increased situational awareness in a city, municipality or even a combat zone. Accurate information regarding the number and location of humans in a particular area is an essential component of situational awareness. Automatically detecting humans by passively detecting skin would be extremely beneficial to law enforcement personnel, soldiers and security officers. This knowledge would provide an enhanced picture of a particular area of interest or operating environment.
At present, several methods employing video surveillance technology are being developed to detect people and/or skin. Typically, these methods exploit either RGB color-matching or spatial object recognition methods to locate humans in a particular scene. Unfortunately, these methods are limited due to a susceptibility to false alarms. RGB or similar color-matching methods rely on low-spectral resolution color bands located in the visible portion of the spectrum and are unable to discriminate between skin and skin-like colored objects, such as can be found in many paints and dyes. In addition, RGB methods are unable to recognize highly pigmented skin.
Other systems, such as those that employ object recognition methods, rely on an ability to associate a shape of an object in a scene with an individual. One limitation associated with this method is that it is necessary for an individual in the scene to span an extended portion of the field-of-view in order to obtain accurate shape identification. In addition, shape matching methods are susceptible to false negatives, such as may occur if shapes found in a scene are not recognized. For example, shape matching methods cannot reliably detect two individuals walking arm-in-arm, or individuals carrying large objects.
It is well known that Hyper Spectral Imaging (HSI), which is defined as many tens or hundreds of narrow spectral bands in either the visible or infrared (IR) portion of the spectrum, is capable of adding significantly to information contained in an image as compared to conventional (three color wide band) imaging. In addition, numerous studies and patents have shown that HSI imaging can detect camouflage, crop variations, provide discrimination of various targets, and potentially identify carcinomas. Unfortunately, conventional HSI sensor systems are data transmission intensive, i.e. require data transmission rates to be several orders of magnitude higher than conventional video systems, and/or computationally intensive, i.e., require the processing of tens of thousands of pixels at many wavelengths simultaneously. Data transmission becomes a problem when processing the spectral image is separated from the data gathering, e.g., during use of an unmanned aerial vehicle, due to power or space limitations or simply due to the complexity of the data that must be processed. Even when located with the sensor, current data processing of all pixels in all bands requires teraflop class processing speeds.
In addition to the above, full HSI sensors are very complex. That is, full HSI sensors must gather data simultaneously in many HSI bands across multiple spatial dimensions. This level of complexity results in a significantly high unit cost. Solutions to this challenge have involved linear scanning arrays which are not well suited to rapidly image large areas, or staring systems which are unable to simultaneously process the hyper-spectral dimension and are thus not well suited to spectrally image moving scenes.
Based on the above, there exists a need for a low cost, effective imaging system that can accurately detect skin from visual images. More specifically, there exists a need for an imaging system that employs HSI technology and utilizes a very limited number of narrow bands that enables certain features in spectral images to be isolated to detect skin. An imaging system of this type would find a wide range of applications, such as military surveillance and reconnaissance systems, as well as facility security and related video tracking systems.